Spinal cord injury (SCI) results in permanent impairment of sensory, motor and autonomic function. Epidural electrical stimulation (EES) applied below the lesion can restore voluntary movement, autonomic function and locomotion following chronic SCI. However, impaired sensation below the SCI does not improve during the application of sublesional EES. Here we present first-in-human results demonstrating simultaneous lower extremity motor activation and somatosensory feedback in three participants with motor complete, chronic SCI enabled by perilesional EES. We determined motor- and sensory-specific EES parameters by leveraging modern deep learning methods and participant-directed control of stimulation. Supralesional EES evoked sensations were synchronized with leg movement, enabling participants to accurately report leg position. We then applied simultaneous supralesional and sublesional EES, enabling intentional control over leg movements and somatosensory feedback during functional tasks. Overall, we demonstrate a perilesional EES framework to modulate sensorimotor function that may improve quality of life in individuals with SCI.

Detecting small extracellular vesicles is critical for understanding disease biology and developing diagnostic tools, yet current methods require lengthy isolation steps and lack sensitivity owing to interference from abundant proteins. Here we report on an assay that uses Janus particles that enable rapid, isolation-free detection by exploiting Brownian rotation-induced blinking changes. When vesicles bind, their size significantly alters the blinking frequency, while smaller proteins produce no signal, ensuring selectivity. Using less than 10 μl of sample, the assay detects approximately 200 vesicles per microlitre and works directly on plasma, serum, urine and cell media in under 1 h. In a blind study of 87 subjects with colorectal cancer, pancreatic ductal adenocarcinoma, glioblastoma, Alzheimer's disease and healthy controls, the method identified disease type with an area under the curve of 0.90-0.99. Compared with ultracentrifugation combined with surface plasmon resonance, which requires 24 h, our approach delivers 2 orders of magnitude better sensitivity and dynamic range, offering a fast and robust platform for clinical and research applications.

Current SARS-CoV-2 (severe acute respiratory syndrome coronavirus 2) vaccines have shown robust induction of neutralizing antibodies and CD4+ T cell activation; however, CD8+ responses are variable, and the duration of immunity and protection against variants are limited. Here we repurpose our DNA origami vaccine nanotechnology DoriVac to target infectious viruses, namely, SARS-CoV-2, HIV and Ebola. The DNA origami nanoparticle, conjugated with infectious-disease-specific heptad repeat 2 peptides, which act as highly conserved antigens, and CpG adjuvant at precise nanoscale spacing, induces neutralizing antibodies, Th1 CD4+ T cells and CD8+ T cells in naive mice, with significant improvement over a bolus control. Pre-clinical studies using lymph-node-on-a-chip systems validate that DoriVac, when conjugated with antigenic peptides or proteins, induces promising cellular and humoral immune responses in human cells. Moreover, DoriVac bearing full-length SARS-CoV-2 spike protein achieves immune responses comparable to current mRNA vaccine platforms while potentially reducing storage constraints. These results suggest that DoriVac holds potential as a versatile, modular vaccine platform, capable of inducing both humoral and cellular immunities, underscoring its potential future use.

Chimeric antigen receptor (CAR) T-cell therapy holds great promise for patients with cancer, and the identification of predictive biomarkers is crucial in finding new ways to guide therapy. Major challenges to the application of informatics and machine learning in CAR T-cell therapy include limited sample sizes and non-uniformity in data generation across cancer indications and trials. Here we took a global, pan-haematologic cancer approach, analysing 256 patients across 5 cancer types and 13 clinical trials. We generated data using a framework that included pre-infusion clinical features, over 2 million apheresis T cells analysed by flow cytometry using 17 unique markers, ex vivo T-cell expansion during CAR T-cell manufacture, more than 90,000 measurements of 30 serum markers and serial tracking of circulating CAR T cells using qPCR. From this data resource, we demonstrate the potential of pan-cancer predictive biomarkers that capture generalizable characteristics of treatment response and non-response in CAR T-cell therapy.

Identifying evolutionarily remote antimicrobial peptides (AMPs) is crucial for discovering underexplored clinical candidates to combat antibiotic resistance. Existing experimental and computational methods are limited by their reliance on sequence identity to known AMPs, missing distant homologues. Here we introduce HMD-AMP, a protein language model-based approach for AMP discovery. HMD-AMP outperforms previous methods in identifying evolutionarily distant AMPs and enables the discovery of unknown and highly potent AMPs from metagenomic data. Applied to host and gut microorganism genomes of nine mammals, HMD-AMP revealed over 37 million predicted AMPs. Of 91 high-confidence sequences experimentally validated, 74 showed strong antibacterial activity and 48 were evolutionarily remote from known AMPs. Four of these AMPs exhibited broad-spectrum antibacterial activity at low effective concentrations and showed low toxicity, with the most potent peptide demonstrating therapeutic efficacy in a mouse model of peritoneal Escherichia coli infection. This study introduces an effective strategy to uncover AMPs.

Post-translationally modified proteins are crucial autoantigens in autoimmune diseases, with citrullinated proteins being key targets of autoantibodies in rheumatoid arthritis (RA). However, accurate citrullinome profiling and autoantigen identification remain limited by insufficient detection methods and computational tools. Here we develop Iseq-Cit (internal standard-assisted enrichment-free approach for high-throughput quantitative analysis of citrullinome), for global citrullinome profiling in individuals at RA risk and in patients with RA across a longitudinal cohort, requiring less than 1% of the sample input needed for conventional methods. We find that plasma citrullinome profiles closely correlate with RA development and severity. Moreover, we develop models integrating clinical indicators and citrullination data, achieving high accuracy in predicting treatment response. To evaluate the RA-sera reactivity of identified citrullinated peptides, we train a bidirectional gated recurrent unit model using 67,399 RA-sera negative and 8,816 RA-sera positive peptides. External validation through enzyme-linked immunosorbent assays confirms 84.2% accuracy in predicting RA-sera reactivity of citrullinated peptides, yielding 19 promising candidates for RA diagnosis. This work provides strategies for citrullinated peptide identification, autoantigen discovery and RA treatment stratification.

Foundation models in artificial intelligence are revolutionizing healthcare by utilizing large-scale unlabelled data for pretraining. However, their intraoperative applications remain underexplored owing to limited surgical data and the challenges of real-time deployment. Here we show the development of the ophthalmic video foundation model (OVFM), designed for microscopic ophthalmic surgical recognition and navigation. Leveraging a self-supervised video transformer structure and trained on an ophthalmic video dataset comprising 1.1 million clips across 144 surgical types, OVFM learns the spatiotemporal motion features of ophthalmic procedures. We demonstrate OVFM's superior performance across seven downstream tasks. To enable real-time use, we applied knowledge distillation, reducing the model's size while retaining its accuracy, which allows for deployment on surgical microscope units. In cataract surgeries performed by ten surgeons on wet-lab porcine eyes, the OVFM-powered system enhanced surgical performance and reduced skill gaps, demonstrating notable potential for real-time, intraoperative applications across various surgical fields.

Metastatic brain disease occurs in up to 30% of patients with lung, melanoma and breast cancers, and the median survival time remains less than a year. Treating these patients is a challenge because surgical approaches are limited and most chemotherapeutic drugs and immunotherapies are ineffective at crossing the blood-brain barrier (BBB). Given the unique abilities of macrophages to cross the BBB and exert their phagocytic function on tumour cells, we genetically engineer macrophages that express a chimaeric antigen receptor (CAR) targeting mesothelin (MSLN). To specifically target metastatic brain tumours, we fused the cells with the immune signalling molecule MyD88. This chimaeric antigen receptor macrophage (CARMA) penetrates the BBB and decreases brain metastasis growth in a humanized mouse model. MSLN-CARMA shows antigen-specific phagocytosis activity against tumour cells and exhibits a bystander effect by releasing TNF to act on surrounding tumour cells lacking the tumour antigen. These features of CARMA represent advantages over other immune therapies and CARMA may serve as a promising therapeutic tool for the treatment of brain metastasis.

Deep learning models that infer clinically relevant biomarker status from tissue images are being explored as rapid and low-cost alternatives to molecular testing. Here we show, through statistical analysis across multiple cancer types, datasets and modelling approaches, that the datasets used to train these models contain strong dependencies between biomarkers and clinicopathological features, which prevent models from isolating the effect of a single biomarker and lead them to learn confounded signals. Consequently, their prediction accuracy varies substantially with the status of codependent biomarkers and clinicopathological variables, and for several biomarkers, the gain over what a pathologist can already infer from routine histopathological features, such as grade, remains modest. These findings indicate that current approaches are not yet suitable as substitutes for molecular testing but can support triage or complementary decision-making with caution. Unconfounded biomarker prediction will require models that learn causal rather than correlational relationships between biomarkers and tissue morphology.